自定义函数替换卷积神经网络reluLayer的matlab代码

以下是一个示例的自定义函数，用于替换matlab中的reluLayer：

function layer = customReluLayer(name)
% 自定义ReLU层
% 输入参数：
%   name：层的名称

layer = nnet.layer.ReLULayer('Name', name);
layer.Description = 'Custom ReLU layer';
layer.InputNames = {'input'};
layer.OutputNames = {'output'};
layer.ForwardFunction = @reluForward;
layer.BackwardFunction = @reluBackward;

end

function Z = reluForward(X)
% 前向传播函数
% 输入参数：
%   X：输入张量
% 输出参数：
%   Z：输出张量

Z = max(0,X);

end

function [dX,dW] = reluBackward(X,~,dZ,~)
% 反向传播函数
% 输入参数：
%   X：输入张量
%   dZ：输出梯度张量
% 输出参数：
%   dX：输入梯度张量
%   dW：权重梯度张量

dX = dZ .* (X > 0);
dW = [];

end

使用此自定义函数可以像使用matlab自带的reluLayer一样使用：

% 创建网络
lgraph = layerGraph();
lgraph = addLayers(lgraph, imageInputLayer([28 28 1],'Name','input'));
lgraph = addLayers(lgraph, customReluLayer('relu'));
lgraph = addLayers(lgraph, fullyConnectedLayer(10,'Name','fc'));
lgraph = addLayers(lgraph, softmaxLayer('Name','softmax'));
lgraph = addLayers(lgraph, classificationLayer('Name','output'));
lgraph = connectLayers(lgraph, 'input', 'relu');
lgraph = connectLayers(lgraph, 'relu', 'fc');
lgraph = connectLayers(lgraph, 'fc', 'softmax');
lgraph = connectLayers(lgraph, 'softmax', 'output');

% 训练网络
options = trainingOptions('sgdm');
net = trainNetwork(trainImages,trainLabels,lgraph,options);

% 测试网络
predLabels = classify(net, testImages);
accuracy = sum(predLabels == testLabels)/numel(testLabels);
fprintf('测试集准确率：%f\n', accuracy);